基于SLIC0超像素分割的阴影检测算法

A Shadow Detection Method Based on SLIC0Superpixel Segmentation

针对单幅图像阴影检测问题,提出了一种基于SLIC0(simple linear iterative clustering zero)超像素分割的阴影检测算法。首先采用SLIC0超像素分割算法对含阴影图像进行分割,生成超像素块并检测出阴影轮廓,然后采用提出的基于融合特征的支持向量机方法,将超像素块分类合并,检测出阴影区域。最后将该算法与Otsu阈值法、传统SVM分类法进行对比,检测结果验证了该算法的有效性。同时,结构相似度(SSIM)与峰值信噪比(PSNR)指标对比表明,该算法较参考算法的检测性能更优。

This paper proposes a shadow detection algorithm based on SLIC0 superpixel segmentation to address the issues of single-image shadow detection.Firstly,SLIC0 superpixel algorithm is used to segment the shadow image to generate superpixel blocks and detect the shadow contour.Then,a feature fusion support vector machine(SVM)method is proposed to classify and merge superpixel blocks to detect the shadow regions.Finally,the proposed algorithm is compared with Otsu threshold method and traditional SVM-based detection method.The experiment results verified the effectiveness of the proposed algorithm.The SSIM and PSNR comparison indicates that the proposed algorithm obtains relative higher performances than the reference algorithms.